2005 tyan motherboard capacitors

[johnson5000]

System (dual cpu Tyan K8WE) would turn on for a few seconds and then turn off. Looked at capacitors and a couple were bulged at the top so replaced them with cheap ebay caps. This worked and it ran fine, so I should have stopped there. I figured that the rest of the caps were going to need replaced eventually so I order polymer caps that matched the originals, about 50 in total of various ratings with Nichicon RME caps. Replaced them all and now the board will turn on for a few seconds and then shut off like before. I read that electrolytics can't always be replaced with polymers so I order some Panasonic FM-A and replace all the 680uf caps which were the ones that were bulging with those, and still same issue. I'm tempted to just undo all the replacements I did and put back the originals. If I wait a while or short out the power pins, the system will turn on for a longer time, about 30 seconds (normally its 2-3 seconds) but no POST or activity except cpu1 getting hot. Does anyone know what the issue could be?

[R_J]

Have you tried a different power supply? Sometimes when the power supply caps go bad, it's the motherboard caps that try to filter the ripple from the power supply and then they also fail.
Also check that you didn't damage any feed thru when removing the caps. Check the original cap leads to see if the feed thru is on the lead.

[johnson5000]

Yes I have tried another power supply and there is no difference. I do not see any feed thru vias on the old caps.

[Topcat]

I'd be willing to wager there was a bunch of bad UCC KZG's on it. KZG series of that era would die on the shelf. They all need to be replaced. I wouldn't replace those with polymer unless you really know what you're doing, these boards are very fickle (they don't respond well to polymodding)....I've dealt with a lot of them back in the day. Modern lytic replacements I use today in place of the KZG are Rubycon ZLH, never had an issue in doing so.

The no-POST issue knowing there's bad caps could be a roasted CPU or northgridge, if it was not post'ing prior to recapping. If it's not posting AFTER the recap, I'd blame the poly mod or possibly through hole damage or a solder bridge somewhere. There are some tight clearances in places on these boards.

[momaka]

Could we see some pictures of your board? Perhaps we may be able to spot something that you haven't.
I don't think the polymods could have killed a MOSFET or a regulator... but if push comes to shove, it may be worthwhile to check all of the major voltage rails for short-circuit.
The board also has nVidia chipsets. While these are Pro chipsets, which aren't exactly as problematic as the consumer nVidia stuff... it's not totally impossible that one of these going finicky just needed a little bit of board flexing/bumping to decide to quit. On that note, see if the board symptoms change if you apply slight pressure to any (or both) of the nVidia chipset heatsinks.
The stock AMD heatsinks are also another point of induced failure - some of them warp the board way too much and can cause CPU socket to develop bad BGA.
If the board can post with 1 CPU only, give that a try and see if it will POST. Also try this both with and without the CPU heatsink clamped down.
And of course, re-inspect all solder joints for shorts or check resistance across all caps that you installed to see if you can find any short-circuits (resistances less than 3 Ohms, usually, depending on multimeter.) Note: it may be helpful to remove the CPU when testing the resistance to rule out false-positive short-circuits.

[johnson5000]

OK I have went through each capacitor replaced on the motherboard and they all appear to be good, no shorts. However, I did notice that pins 11(+12V) and 12(+3.3V) short to pin 24(Ground) on the ATX power connector so I am not sure how that has happened. Here are some photos.

[Per Hansson]

The short is probably in a MOSFET or regulator.
If your meter shows fractions of an ohm you can sometimes find it by just looking for a lower and lower ohms value on the meter as you approach the faulty component.
Or you could use a thermal camera to see if something gets really hot when the board does run...

[R_J]

Closely inspect the positive capacitor leas where they go through the board, sometimes the clearance between where the lead comes through the board and the ground is very small and there could be a short.

[johnson5000]

Originally Posted by R_J

Closely inspect the positive capacitor leas where they go through the board, sometimes the clearance between where the lead comes through the board and the ground is very small and there could be a short.

Do you mean shorting with the case? I am using the board outside of the case on a table so that shouldn't be the issue.

[pfrcom]

Originally Posted by R_J

Closely inspect the positive capacitor leas where they go through the board, sometimes the clearance between where the lead comes through the board and the ground is very small and there could be a short.

My similar experience:

https://badcaps.net/forum/showpost.p...3&postcount=16

https://badcaps.net/forum/showpost.p...4&postcount=17

[R_J]

Originally Posted by johnson5000

Do you mean shorting with the case? I am using the board outside of the case on a table so that shouldn't be the issue.

This picture shows where the positive lead cones through the board, as you can see the ground circuit is very close to where the lead is, and it would not take much to short it to ground.
Your board may be different but I just thought I would mention it.

[johnson5000]

OK I see what you mean. I don't see any of these, but I tested the atx pin 24 ground with the positive leads on all the caps and a few of them are shorted, is this some kind of multilayer pcb via short? Tried removing one of the caps and wicking up all the solder I could but it is still shorted, not sure how this can even be fixed?

[johnson5000]

Tried hooking up a dc power supply and using a flir thermal camera and nothing was getting hot. Connected it to a atx power supply and pressing on a bunch of times and one of the voltage regulators near the pci slots got slightly warm on the flir but it doesn't seem shorted or burnt out or anything?

[momaka]

Originally Posted by johnson5000

However, I did notice that pins 11(+12V) and 12(+3.3V) short to pin 24(Ground) on the ATX power connector so I am not sure how that has happened.

Likely a fluke. I find it unlikely that you would have a short-circuit on two separate voltage rails at the same time (unless you did a really terrible job with the soldering... which also seems unlikely.)

Moreover, if you had a true short-circuit, the PSU would not start (usually the fans and power will flicker On for less than a second.)

I suspect you used "continuity" test mode on your multimeter and heard a beep, hence assuming the short-circuit. If yes, I suggest not to use this method. Always measure the resistance instead, starting with lowest resistance setting on the multimeter (if using a manual-ranging one.) If this is what you did, post what results (resistances) you got here.

Also, if PSU was connected to the motherboard when you tested it, some PSUs have low-value resistors on the output rails as "dummy" loads. This can sometime trick you into thinking there is a short-circuit when in fact there isn't. For this reason, only measure resistances with the PSU disconnected. In some instances, even the CMOS battery can interfere with resistance measurements... though that's usually only when doing such near the SIO/LPC chip.

Lastly, keep in mind that caps can charge up... so if you tried to measure resistance/continuity on a rail with many caps and then reverse the probes and try to measure again, the multimeter may show a short-circuit / low resistance temporarily, until the caps charge in the reverse direction. So for cases like that, only take note of the resistance after it has settled down to a steady value (usually >3 seconds.)

Anyways, since the motherboard originally had the issue of not turning on (or rather, only running for a few moments) and now you're back at it after the recap, I suggest trying some of the more common PC troubleshooting techniques:
- re-seat CPU and RAM a few times
- try with only 1 CPU and 1 RAM stick
- try with and without fully clamping the CPU heatsink - i.e. once with the CPU HS clamped properly, and once without it being clamped but only sitting on the CPU... but in both cases with thermal compound(!)
- reset CMOS (remove CMOS battery / set jumper to reset config / etc.)
- make sure CPU heatsink fan is connected to the board. Generally, this was only applicable to old AMD board from the socket 462 era, some of which would not boot without the CPU fan connected to the motherboard (to hopefully prevent damage to the CPU.) I don't think that would be the case here with this motherboard, but worth a try anyways.
- remove all RAM and see if the motherboard powers on and beeps an error code (may need to connect an external speaker/buzzer if board doesn't have one.) If it does power On and stays On, you may be having a memory-related error (though not 100% necessarily.) If not, you may be having a power issue somewhere.
- for the few moments that the motherboard does turn On, see if you can measure the voltage on common power rails on the motherboard, like CPU V_core, CPU V_tt, RAM V_dd, RAM V_tt, Northbridge & Southbridge Vcc rails (there may be more than 1), PCI-E comm, and voltage on any linear regulators.

[johnson5000]

The resistance on the 3.3 and 12 atx pins are 0.1 ohms and when i connect it to the dc power supply it says its shorted also so pretty sure it is shorted. I have tried all of what you said multiple times and just cant figure it out. I just decided to get an asus motherboard of similar specs on ebay for $50 and forget about this tyan board. This system is obsolete anyways and just wanted it working for sentimental values since it has a custom water cooling system and case that looks neat when it is on. I can mail this tyan board to you if you are curious and want to see for yourself but otherwise I think it is going to go on the shelf.